Relabeling in napari workflows

[mfranzon]

Close #138
Ref #172
Close #180

[jluethi]

I had a brief look at the 314 implementation. This assumes labels will always be ordered and the last entry will have the highest value. Can we make this assumption? I think it holds up most of the time for our examples. But we don't do any sorting, so there may be napari workflows a user can run that break this assumption.
Can we switch to a max() based increase of the label values?

Also, I'm wondering for this approach: We adapt the label values in the table here independently of how we'd adjust the ones in the images. It is of utmost importance that the matching remains 100% correct, i.e. we can't have any cases where after relabeling, image 2 has label 731 but the label 731 entry in the dataframe is from a measurement on image 3. How are we ensuring this does not happen?